## _wp1515 - References

## Source details

**Canonical URL:** [_wp1515 - References](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1515.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1515.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1515.pdf.json)

---

### I. Introduction and Regional Context
- Common credit cycle in CESEE:
  - Credit expanded very rapidly over 2003–08 — credit-to-GDP ratio increased on average by some 30 percentage points.
  - As the global financial crisis struck, credit growth stalled or became negative; credit-to-GDP ratio fell or stabilized in 2009–12.
  - Turkey is the only notable exception.
- Drivers of the boom and bust:
  - Boom drivers: convergence and integration with the rest of Europe; presence of Western banks with easy access to liquidity; low real interest rates; domestic demand boom; buildup of macroeconomic and financial imbalances.
  - Bust drivers: economic slowdown, heightened uncertainty, withdrawal of funding by parent banks; over-indebted households, corporates, and banks; banks tightened lending policies.
- Cross-country heterogeneity:
  - Parent bank funding central to boom in the Baltic countries.
  - Countries with greater reliance on domestic funding and/or flexible exchange rates avoided more imbalances.

### Box 1 — Common Factors of the Boom-Bust Credit Cycle (summary)
- Boom (demand and supply factors):
  - Demand-side: strong economic growth, rapid income convergence, rising house prices, low real interest rates, and in some countries pro-cyclical fiscal policies.
  - Supply-side: unprecedented inflows of external funding; foreign-owned subsidiaries with access to cheap parent bank funding; perception of under-banked markets; lax lending standards.
- Post-2008/09 adjustments:
  - Demand fell: recessions, lower incomes, house price drops, exchange rate adjustments affecting un-hedged borrowers.
  - Supply fell: decline in global risk appetite, reversal of capital flows, banks building capital/liquidity buffers, rising NPLs, tighter lending standards, higher funding costs and lower interest margins.
- Identification challenges:
  - Credit demand and supply are unobservable; some factors affect both.
  - Matched bank-firm datasets improve identification but are unavailable for CESEE in this paper.

### II. Research Approach — Two Complementary Methods
- Objective: analyze role of demand and supply factors in explaining recent credit cycle in CESEE.
- Two complementary approaches:
  - Cross-country panel study:
    - Bank-level panel dataset for CESEE; period 2001–11.
    - Sample: more than 400 banks from 20 countries (all CESEE except Russia).
    - Annual real credit growth regressed on lagged bank characteristics, contemporaneous domestic macro variables, and contemporaneous EMBIG spread; include crisis and foreign-ownership dummies and interactions.
  - Country case studies:
    - Disequilibrium model jointly estimating credit demand and supply.
    - Five case studies: Latvia, Lithuania, Montenegro, Poland, and Romania.
    - Uses new lending flows (real terms) and estimates whether demand or supply constrained actual lending (quantity rationing).

### Key Findings (summary)
- Both demand and supply factors explain evolution of credit growth; relative importance shifted over time and across countries.
- Panel-average (CESEE):
  - Both demand and supply matter.
  - Sensitivity of credit growth to supply factors increased post-crisis.
  - Sensitivity to demand factors (particularly macroeconomic factors) decreased.
  - Relative importance of supply factors increased after the crisis.
- Case-study results:
  - Both credit supply and credit demand rose during the boom and fell during the bust.
  - Supply-side constraints became more important in the late-crisis period in some countries (Lithuania and Montenegro).
  - Heterogeneous country experiences reflect intensity of boom, availability of funding, depth of output collapse, and country-specific policies.

### Literature and Background (selected themes)
- Topics covered in literature: foreign bank entry and performance; fast credit growth as financial deepening vs excessive growth; role of international capital flows and parent funding; loan demand and discouragement.
- References cited include studies by Claessens et al., Bakker and Gulde, Lane and McQuade, Mendoza and Terrones, IMF reports, and others.

### III. Regional Bank-Level Panel Analysis — Data and Estimation
- Data and sample:
  - Bankscope bank-level data matched with time-varying bank ownership from Impavido, Vandenbussche, and Zeng (2015).
  - Sample: more than 400 banks from 20 countries (2001–11); Russia excluded.
  - Annual data matched with macro variables from WEO and global variables from Bloomberg.
  - Outliers: banks in top and bottom 1 percent of distribution for each bank-level variable (except bank size) dropped.
- Estimation strategy:
  - Panel fixed effects estimator with three steps:
    1. Regress annual real credit growth (local currency) on lagged bank characteristics, contemporaneous domestic macro variables, and EMBIG spread; control for 2008/09 crisis and bank ownership.
    2. Interact macro and bank variables with a crisis dummy.
    3. Interact macro and bank variables with a dummy capturing foreign ownership and crisis; supplement with parent bank characteristics.
- Variable treatment and interpretation:
  - Dependent: real growth of gross loans in local currency.
  - Macroeconomic regressors: real domestic demand growth (+), average inflation (-), EMBIG spread (-).
  - Bank regressors (lagged): bank assets-to-GDP (-), loan-loss reserves as percent of gross loans (-), net loans as percent of customer deposits (-), liquid assets to deposits and short-term funding (+), equity to net loans (+), return on average equity (+).
  - Foreign ownership defined: foreign global-ultimate-owner controls 25 percent or more of total shares; ownership can change over time.
  - Parent bank characteristics used in alternative specifications: parent bank home country CDS spreads (-), lagged parent equity to total assets (+).
  - Modeling caveat: panel regressions tag bank characteristics as supply changes; macro variables mostly tagged as demand factors.

### Main Empirical Findings from Panel Regression
- Both demand and supply factors significant in explaining credit growth.
- Pre- vs post-crisis time variation:
  - Coefficients on bank-specific variables (reserves to gross loans ratio, net loan to deposit ratio, return on equity) increased in size after the crisis.
  - Domestic demand coefficient fell: pre-crisis (2001–07) a one percentage point rise in domestic demand associated with a 1.8 percentage points increase in banks’ annual real credit growth; during 2008–11 a similar shock associated with an impact less than half that size (interaction estimates: × crisis dummy -1.252*** (0.235) in Column (2)).
  - Impact of inflation on credit remained stable across the financial cycle.
  - Post-crisis: solvency became more important; liquidity coefficient reduced while solvency coefficient increased significantly.
- Foreign ownership findings:
  - Foreign dummy large and highly significant (e.g., Column (1): 19.511*** (3.648); Column (2): 19.965*** (3.578)).
  - Effect of foreign ownership diminished over crisis years.
  - Domestic banks grew faster with higher return on equity; for foreign banks this holds primarily during the crisis period.
  - Foreign banks’ credit expansion more sensitive to solvency indicators (equity to net loan ratio) than domestic banks’.
  - Parent bank characteristics matter: parent bank home CDS spread negative (Column (3): -0.037*** (0.010)); parent bank equity to net loan ratio positive (Column (3): 2.479*** (0.435)).
  - Average equity to net loan ratios for domestic banks were higher than for foreign banks pre- and post-crisis.
- Selected coefficient estimates (CESEE: Determinants of Credit Growth, 2001–11) — preserving exact figures:
  - EMBIG spread:
    - Column (1): -0.014*** (0.005)
    - Column (2): -0.019*** (0.005)
    - Column (3): -0.024*** (0.006)
  - Real domestic demand growth:
    - Column (1): 0.977*** (0.100)
    - Column (2): 1.794*** (0.199)
    - Column (3): 1.544*** (0.194)
    - × crisis dummy (Column (2)): -1.252*** (0.235)
    - × crisis dummy (Column (3)): -0.865*** (0.222)
  - Average inflation:
    - Column (1): -0.806*** (0.119)
    - Column (2): -0.754*** (0.119)
    - Column (3): -0.656*** (0.117)
  - Bank size to GDP (first lag):
    - Column (1): -2.911*** (0.428)
    - Column (2): -3.804*** (0.530)
    - Column (3): -3.146*** (0.498)
    - × crisis dummy (Column (2)): 0.852*** (0.263)
  - Reserves to gross loans ratio (first lag):
    - Column (1): -1.238*** (0.209)
    - Column (2): -0.629*** (0.219)
    - × crisis dummy (Column (2)): -0.763** (0.346)
    - × crisis dummy (Column (3)): -1.077*** (0.403)
    - × foreign dummy (Column (3)): -1.785*** (0.440)
    - × crisis × foreign dummy (Column (3)): 1.373** (0.534)
  - Equity to net loan ratio (first lag):
    - Column (1): 0.497*** (0.059)
    - Column (2): 0.428*** (0.060)
    - Column (3): 0.402*** (0.066)
    - × crisis dummy (Column (2)): 0.626*** (0.106)
    - × crisis dummy (Column (3)): 0.309*** (0.087)
    - × foreign dummy (Column (3)): 0.643*** (0.119)
  - Liquid assets to deposits and st funding ratio (first lag):
    - Column (1): 0.212*** (0.053)
    - Column (2): 0.258*** (0.059)
    - Column (3): 0.219*** (0.054)
    - × crisis dummy (Column (2)): -0.222*** (0.078)
  - Return on equity (first lag):
    - Column (1): 0.207*** (0.058)
    - Column (2): 0.163* (0.085)
    - × crisis dummy (Column (2)): 0.165** (0.074)
    - × foreign dummy (Column (3)): -0.410*** (0.157)
    - × crisis × foreign dummy (Column (3)): 0.587*** (0.149)
  - Foreign dummy:
    - Column (1): 19.511*** (3.648)
    - Column (2): 19.965*** (3.578)
- Sample and fit statistics:
  - Observations:
    - Columns (1) and (2): 2,415
    - Column (3): 2,093
  - R-squared:
    - Column (1): 0.336
    - Column (2): 0.372
    - Column (3): 0.404
  - Number of banks:
    - Columns (1) and (2): 435
    - Column (3): 415
  - Bank fixed effects: yes
  - Sources: Bankscope, Bloomberg, WEO database, and authors' calculations.

### IV. Disequilibrium Model — Estimation and Implementation (Box 3)
- Model structure:
  - Underlying model: interest rate not perfectly flexible; observed new lending Ct = min(estimated demand, estimated supply).
  - Explanatory vectors include interest rate and non-price determinants; error terms jointly normal, independent over time, zero mean, covariance ∑.
- Estimation:
  - Maximum likelihood method (Maddala and Nelson, 1974).
  - Switching regression sensitive to specification; insignificant variables dropped except interest rate.
  - Statistical significance of estimated excess supply assessed via Monte Carlo simulations (typically 5,000 repetitions) to generate confidence intervals; fixed-error confidence bands reported and time-varying bands in Appendix V.
- Dependent variable and data treatment:
  - Dependent: new loans extended (flow), real terms; includes new loan contracts and rollovers.
  - Monthly flows or 3- or 6-month smoothed averages used; smoothing may introduce potential endogeneity concerns but reduces noise.
  - For some countries regressions estimated separately for households and NFCs when data allow.
- Choice of explanatory variables (a priori exclusion restrictions preserved):
  - Demand variables: real lending rate (-), inflation expectations (+), confidence surveys (+), stock market changes (+), real GDP/retail/industrial production (+), volatility of consensus forecasts (-), debt overhang measures (-), profitability proxies (+).
  - Supply variables: real lending rate (+), interest margin (+), GDP growth (+), collateral values (+), NPL ratio (-), funding costs (-), deposits/parent funding (+), banking system capital (+).
- Treatment of parent funding endogeneity:
  - Two-step estimation: first stage regress parent funding on instruments (parent bank or home country CDS spreads); second stage include fitted values in supply equation.
  - Fitted parent-funding variable captures variation related to parent bank health/stress; significance must be narrowly interpreted.
- Model performance and general findings:
  - Predicted and actual credit are fairly close; most coefficients have expected signs but interest rate sometimes not significant.
  - Both demand and supply rose during the boom and fell in the bust across case studies.
  - Country heterogeneity in timing and extent of demand vs supply constraints:
    - Montenegro and Lithuania: supply became more constraining post-crisis.
    - Latvia: tightening supply and demand contributed broadly equally to contraction.
    - Poland and Romania: demand factors constrained credit most of the time.

### V. Case Studies — Key Country Findings and Statistics

- Montenegro (selected chronology and findings):
  - Supply constraints matter most in explaining credit expansion throughout sample.
  - Parent bank funding increased from 7 to 27 percent of GDP between 2006 and 2008.
  - NPLs exceeded 7 percent by end-2008 (above the 4 percent average of the other four case-study countries).
  - Credit-to-GDP ratio increased from 38 to 89 percent of GDP between 2006 and 2008.
  - Parent bank funding fell from 27 to just over 10 percent of GDP between 2008 and 2012.
  - Credit growth was negative for four consecutive years (2009–12).
  - Credit-to-GDP ratio fell by some 34 percentage points of GDP over 2009–12 (largest decline among case studies).
  - Loan-to-deposit ratio peaked in 2008 at 141 percent.
  - Central bank measures included broadening/tightening reserve requirement calculations in 2006 and 2007–08; increasing capital adequacy ratio from 8 to 10 or 12 percent for banks with credit growth in excess of 60 or 100 percent respectively; capping annual credit growth at 30, 40, or 60 percent depending on bank size; later reductions in effective reserve requirement rates and reduced interest rate for reserve requirements for liquidity.
  - Box 4 estimation performance:
    - Demand: interest rate negative and significant; demand positively correlated with real activity; negatively with debt overhang.
    - Supply: deposit growth significantly and positively associated with credit expansion; NPL ratio negatively associated with credit supply.
    - Parent funding (instrumented by sovereign CDS) significant and positive in supply equation (coefficient lower than on deposit funding).
  - Sample period: 2007M1–2012M12.
  - Box 5 contextual stats (selected exact figures):
    - GDP per capita (US$ PPP, 2012): 11,800
    - Population (2012): 622,000
    - During 2004–08, GDP growth averaged 7 percent.
    - Credit-to-GDP ratio increased from 38 to 89 percent of GDP between 2006 and 2008.
    - Parent bank funding increased from 7 to 27 percent of GDP between 2006 and 2008.

- Lithuania (selected findings — Box 6 and Appendix III)
  - Joint estimation for private sector (households and NFCs jointly).
  - Demand equation: profit margins significantly and positively related to credit demand.
  - Supply equation: NPL ratio strongly and negatively associated with credit supply.
  - Parent funding: coefficient on parent funding (instrumented by parent bank CDS) much larger than coefficient on deposits — greater responsiveness of credit supply to parent funding.
  - Dynamics since bust:
    - Demand and supply fell strongly after 2008/09; recovery has not been matched by recovery in credit demand or supply.
    - Two short periods of statistically significant imbalances:
      - Excess supply in 2009Q2–Q3 (small).
      - From early 2012 onward credit demand somewhat exceeds credit supply (consistent with panel findings).
    - Early 2012 onward coincides with significant excess liquidity in banking system; NPL resolution and strict lending standards may explain supply constraints despite excess liquidity.
  - Sample period: 2006M10–2012M10.

- Poland (selected findings — Box 10 and Appendix III)
  - Model estimated separately for households and NFCs.
  - Household credit expanded strongly; mortgages in foreign exchange were prevalent.
  - Most variables significant except:
    - Lending rate in supply equation for households (insignificant).
    - Lending rates and deposit rates for NFCs (insignificant).
  - NPL ratio significant in both household and NFC supply equations.
  - Debt overhang variables have larger coefficients for households than corporates.
  - Parent funding (instrumented by CDS) significant in supply equation for household credit; first-step OLS R square very low.
  - Sample period: 2005M12–2012M9.

- Romania and Latvia (selected diagnostics from Appendices)
  - Romania: demand factors constrained credit most of the time; lending rate to corporate negative and significant in demand; supply-side bank capital divided by minimum capital requirements positive and significant.
  - Latvia: tightening supply and demand both contributed to contraction; demand-side indicators (economic sentiment, industry new orders) positive and significant; NPLs negative and significant in demand and supply.

### VI. Appendices — Data, Fit, Robustness, and Summary Statistics (selected exact figures)
- Appendix I — Panel data summary statistics (selected exact figures):
  - Number of observations: 2,415.
  - Growth of gross loans (%) — No. of observations 2415 — Mean 21.93 — Standard deviation 37.1 — Min -54.2 — Max 345.5.
  - EMBIG spread (pp) — 2415 — Mean 392.1 — Std dev 164.4 — Min 19.7 — Max 1796.4.
  - Real domestic demand growth (percent) — 2415 — Mean 4.1 — Std dev 8.1 — Min -27.4 — Max 24.2.
  - Average inflation (percent) — 2415 — Mean 6.5 — Std dev 7.3 — Min -1.2 — Max 80.6.
  - Bank size (% of host country GDP, 1st lag) — 2415 — Mean 3.8 — Std dev 5.5 — Min 0.0 — Max 43.4.
  - Reserves to gross loan ratio (%, 1st lag) — 2415 — Mean 5.6 — Std dev 5.2 — Min 0.0 — Max 41.9.
  - Net loans to customer deposits ratio (%, 1st lag) — 2415 — Mean 109.0 — Std dev 74.9 — Min 12.9 — Max 679.1.
  - Liquidity to dep. & st funding ratio (%, 1st lag) — 2415 — Mean 38.2 — Std dev 22.9 — Min 1.3 — Max 246.8.
  - Equity to net loans ratio (%, 1st lag) — 2415 — Mean 27.3 — Std dev 21.7 — Min 1.7 — Max 215.7.
  - Return on average equity (%, 1st lag) — 2415 — Mean 8.4 — Std dev 15.1 — Min -99.1 — Max 75.7.
  - Parent equity to total assets ratio (%) — 988 — Mean 6.3 — Std dev 4.9 — Min -85.8 — Max 40.5.
  - Parent bank home country CDS spreads — 988 — Mean 99.8 — Std dev 134.5 — Min 0.0 — Max 812.4.
- Appendix II — Case study summary indicators (selected exact figures):
  - Average real GDP growth (2004-08 / 2009-12):
    - Latvia: 7.0 / -0.1
    - Lithuania: 7.1 / -1.2
    - Montenegro: 7.3 / -1.9
    - Poland: 5.4 / 3.0
    - Romania: 6.8 / -1.9
  - Selected credit stock (percent of GDP) examples:
    - Latvia 2006 38.5, 2008 89.3, 2012 56.3
    - Lithuania 2006 21.0, 2008 60.9, 2012 48.3
    - Montenegro 2006 47.1, 2008 103.1, 2012 75.6
    - Poland 2006 26.6, 2008 47.2, 2012 50.1
    - Romania 2006 17.7, 2008 41.8, 2012 15.9
- Appendix III — Disequilibrium model variable treatment (selected exact elements):
  - Dependent variable: new credit flow in real terms (monthly data, in logs), specific construction and smoothing varies by country (six month moving average for Montenegro and Lithuania; 3-month for Poland; country-specific deflation using HICP).
  - Sample periods:
    - Montenegro: 2007M1–2012M12
    - Lithuania: 2006M10–2012M10
    - Latvia: 2004M12–2012M9
    - Poland: 2005M12–2012M9
    - Romania: 2005M1–2012M8
- Appendix IV and V — Model fit and robustness:
  - Predicted and actual credit series plotted for each case; robustness of excess supply assessed with (+2 std, -2 std) bands using Monte Carlo simulations.

* _wp1515 - References (PDF chapter/section) — content from the supplied source PDF file.*

### References .............................................................................................................

### _wp1515 - References

### I. Introduction and Regional Context
- Countries in Central, Eastern, and Southeastern Europe (CESEE) experienced a common credit cycle:
  - Credit expanded very rapidly over 2003–08 — during this period the credit-to-GDP ratio increased on average by some 30 percentage points.
  - As the global financial crisis struck, credit growth stalled or became negative; credit-to-GDP ratio fell or stabilized in the period 2009–12.
  - Turkey is the only notable exception.
- Drivers of the boom and bust:
  - Boom: convergence and integration with the rest of Europe; presence of Western banks with easy access to liquidity; low real interest rates; domestic demand boom; buildup of macroeconomic and financial imbalances.
  - Bust: economic slowdown, heightened uncertainty, withdrawal of funding by parent banks; over-indebted households, corporates, and banks; banks tightened lending policies in response to perceived excessive credit extension.
- Cross-country heterogeneity noted:
  - Parent bank funding central to boom in the Baltic countries.
  - Countries with greater reliance on domestic funding and/or flexible exchange rates avoided more imbalances (see IMF 2012).

### Box 1 — Common Factors of the Boom-Bust Credit Cycle
- During the boom:
  - Demand-side drivers: strong economic growth, rapid income convergence, rising house prices, low real interest rates, and in some countries pro-cyclical fiscal policies.
  - Supply-side drivers: unprecedented inflows of external funding; foreign-owned subsidiaries with access to cheap parent bank funding; perception of under-banked markets and favorable profit opportunities; lax lending standards.
- After 2008/09 crisis:
  - Demand fell due to recessions, lower incomes, house price drops, and exchange rate adjustments affecting un-hedged borrowers.
  - Supply fell due to decline in global risk appetite, reversal of capital flows, banks building capital/liquidity buffers, rising NPLs, tighter lending standards, higher funding costs and lower interest margins.
- Identification challenges in disentangling demand vs. supply:
  - Credit demand and supply are unobservable; some factors drive both.
  - Recent progress uses matched bank-firm lending datasets, but such rich data are unavailable for CESEE in this paper.

### II. Research Approach — Two Complementary Methods
- Objective: analyze role of demand and supply factors in explaining recent credit cycle in CESEE to inform policy responses.
- Two complementary approaches:
  - Cross-country panel study (Section II):
    - Bank-level panel dataset for CESEE to analyze credit growth using bank-specific and macroeconomic variables.
    - Focus on how relative role of demand and supply changed after the crisis and varied by bank ownership status.
  - Country case studies (Section III):
    - Disequilibrium model of credit demand and supply to estimate demand and supply themselves rather than actual credit growth.
    - Five country case studies: Latvia, Lithuania, Montenegro, Poland, and Romania.
    - Determines, for each country and time, whether credit demand or credit supply constrained credit growth (quantity rationing).

### Key Findings (Summary)
- Both demand and supply factors explain evolution of credit growth; relative importance shifted over time and across countries.
- Panel results (average for CESEE):
  - Both demand and supply factors matter in explaining credit growth.
  - Sensitivity of credit growth to supply factors increased post-crisis.
  - Sensitivity of credit growth to demand factors (particularly macroeconomic factors) decreased.
  - Implies relative importance of supply factors increased after the crisis.
- Case study results:
  - Both credit supply and credit demand rose during the boom and fell during the bust.
  - Supply-side constraints became more important in the late-crisis period in some countries (Lithuania and Montenegro).
  - Heterogeneous country experiences reflect country-specific circumstances (intensity of the boom, availability of funding, depth of output collapse, etc.).
  - Country-specific macroprudential policies may contribute to heterogeneity but this paper does not focus on their role in explaining credit developments.

### Literature and Background
- Early literature focused on privatization and foreign bank entry impacts on banking system performance and credit allocation (Claessens and others, 2001; Bonin and others, 2005; Haas and Lelyveld, 2006; Havrylchyk and Jurzyk, 2010; Aydin, 2008; Degryse and others, 2009).
- Debate on fast credit growth as financial deepening vs. excessive growth presaging booms/busts (Cottarelli and others, 2003; Hilbers and others, 2005; Duenwald and others, 2005; Égert and others, 2006; Enoch and Ötker-Robe, 2007; Eichengreen and Steiner, 2008; Tressel and Detragiache, 2008).
- Role of international capital flows and parent funding highlighted in Bakker and Gulde (2010), Lane and McQuade (2012), Mendoza and Terrones (2012), IMF (2013a).
- Loan demand and discouragement examined for CESEE in Brown and others (2012).

### III. Regional Bank-Level Panel Analysis — Data and Estimation
- Data and sample:
  - Bank-level data from Bankscope on credit growth and bank financial variables matched with time-varying bank ownership from Impavido, Vandenbussche, and Zeng (2015).
  - Sample includes more than 400 banks from 20 countries (all countries in the CESEE region except Russia).
  - Period covered: 2001–11.
  - Annual data matched with macroeconomic variables from WEO database and global financial variables from Bloomberg.
  - Banks in top and bottom 1 percent of distribution of each bank-level variable (except bank size) treated as outliers and dropped.
  - Basic descriptive summary statistics and data coverage are in Appendix I.
- Estimation strategy:
  - Panel fixed effects estimator.
  - Three steps:
    1. Regress annual real credit growth (local currency) of bank i at time t on lagged bank characteristics, contemporaneous domestic macro variables, and contemporaneous EMBIG spread; control for 2008/09 crisis and bank ownership dummies.
    2. Interact macroeconomic and bank variables with a crisis dummy to examine changes during crisis.
    3. Interact macroeconomic and bank variables with a dummy capturing both foreign ownership and the crisis; supplement regressors with parent bank characteristics.
- Russia excluded for two reasons:
  - Raw dataset contained more than half Russian banks; inclusion would likely drive CESEE results.
  - Data on customer deposits for Russian banks in Bankscope use a different definition from other CESEE countries.

### Box 2 — Estimation and Model Specification of Panel Regression
- Dependent variable:
  - Real growth of gross loans in bank i, country j, at time t, expressed in local currency.
- Regressors include:
  - Bank-specific fixed effects ci.
  - Macroeconomic variables at time t in country j: macroj,t.
  - EMBIG spread at time t.
  - Bank-specific financial variables at time t-1: banki,t-1.
  - Dummy for crisis years 2008–11: Dcrisis.
  - Dummy for foreign-owned banks: Dforeign (bank considered foreign-owned if a foreign global-ultimate-owner controls 25 percent or more of total shares; ownership can change over time).
  - Interaction term Dcrisis * Dfor to capture joint effect of crisis and foreign ownership.
- Macroeconomic and bank-specific explanatory variables and expected signs:
  - Domestic macroeconomic variables:
    - Real domestic demand growth (+): used instead of GDP; contemporaneous and treated as exogenous from individual bank perspective.
    - Average inflation (-): captures internal imbalances and monetary policy credibility; expected negative sign.
    - Exchange rate: found insignificant and dropped due to insignificance and high correlation (0.5 percent) with inflation.
  - Global variable:
    - EMBIG spread (-): contemporaneous EMBIG spread used; expected negative correlation with credit growth.
  - Bank variables (lagged one period):
    - Bank assets-to-GDP (bank size to GDP) (-): larger banks expected to grow more slowly.
    - Loan-loss reserves as percent of gross loans (reserves to gross loans ratio) (-): indicates poor asset quality; expected negative effect.
    - Net loans as percent of customer deposits (net loan to deposit ratio) (-): indicates degree of financial leverage; expected negative effect.
- Notes:
  - Credit measured in domestic currency (in euros for Kosovo and Montenegro).
  - Short panel dimension precluded use of mean group estimators or co-integration analysis.
  - Banking sector competition not included due to measurement difficulties.

*Italic source attribution: _wp1515 - References (PDF chapter/section) — content from the supplied source PDF file.*

### Box 2. Estimation and Model Specification of Panel Regression (continued)

### Box 2. Estimation and Model Specification of Panel Regression (continued)

### Variables and expected signs
- Liquid assets as a percent of the sum of customer deposits and short-term funding (liquid assets to deposits and st funding ratio) (+). Higher available liquidity (in the preceding period) is expected to facilitate greater credit expansion.
- Equity as a percent of net loans (equity to net loan ratio) (+), which measures solvency or capital adequacy. Better capitalized banks are expected to be less constrained in their ability to expand credit.
- Return on average equity (return on equity) (+), indicating bank profitability. More profitable banks are expected to be in a better position to extend credit.
- Robustness checks: variables included in the reported specification are only those whose effects remained significant and carried the same sign across all checks.

### Subsequent model specifications and extensions
- Year dummies and interaction terms:
  - Instead of a single crisis dummy, dummies for each year of the crisis are included.
  - Principal variables such as macro j,t and bank i,t-1 are interacted with a dummy for the crisis period to allow effects to vary during boom and bust periods.
- Impact of foreign ownership:
  - Focus on difference between domestically-owned banks and subsidiaries/branches of foreign banks via a foreign ownership dummy.
  - Foreign ownership is restricted to banks owned by foreign legal entities that are banks; banks owned by foreign legal entities other than banks or foreign natural persons are excluded.
  - Rationale: subsidiaries/branches of foreign banks may be integrated into cross-border banking group strategies (funding, capital, liquidity).
  - The foreign ownership dummy is first interacted with a dummy for each crisis year to capture changes over crisis years.
- Parent bank characteristics (replacing foreign dummy in alternative specifications):
  - Parent bank home country CDS spreads to proxy for parent bank funding costs (-).
  - Lagged parent bank equity to total assets ratio as proxy for capital strength (+), along with its interaction with a crisis dummy.
- Dropping insignificant variables:
  - Some bank characteristics, interaction terms, and alternative parent bank characteristics were dropped as insignificant.
  - Specifically tried (lagged) ratio of parent bank net loans to deposits and parent bank cost-to-income ratio; these were insignificant.
- Data constraint note:
  - Data constraints prevented inclusion of a bank-specific cost of funding variable. For subsidiaries of larger banking groups, the CDS spread of the home country was used as a proxy for parent bank funding costs. This proxy allowed greater data coverage and yielded qualitatively similar results as using either the parent bank CDS when available, or the average of the CDS of the three largest parents of the same home country.

### Interpretation: demand versus supply factors
- Tagging variables:
  - Domestic macroeconomic variables (e.g., growth in domestic demand) are assumed here to mostly reflect demand factors, though they can affect both supply and demand.
  - The EMBIG spread is assumed to reflect both demand and supply factors; it is not tagged to either side.
  - Individual bank characteristics are assumed to reflect supply factors (banks' capacity or willingness to lend).
  - Foreign ownership variables are considered a very specific supply factor (facilitating access to foreign funding).
- Modeling caveat:
  - The panel regressions treat variation in bank characteristics as capturing credit supply changes.

### Main empirical findings (overview)
- Both demand and supply factors played a role in explaining credit growth.
- Basic regression (first column in Table 1):
  - Domestic demand and inflation are both significant: domestic demand positively correlated with credit growth; inflation negatively correlated.
  - EMBIG spread is significant and negative.
  - On the supply side, banks expanded lending more rapidly when:
    - they were smaller;
    - their asset quality was better;
    - their solvency was higher;
    - they were more liquid.
- Time variation (interaction with crisis dummy; second column in Table 1):
  - After the crisis, coefficients on bank-specific variables (reserves to gross loans ratio, net loan to deposit ratio, return on equity) increased in size.
  - Interpretation: given equal fundamentals, banks extended less credit after the crisis than before.
  - Pre-crisis: better liquidity and capital adequacy had positive effects on credit growth.
  - Post-crisis: solvency became much more important; liquidity coefficient was reduced while solvency coefficient increased significantly.
  - Domestic demand coefficient fell: during 2001–07, a one percentage point in domestic demand was associated with a 1.8 percentage points increase in banks’ annual real credit growth; during 2008–11, a similar shock is associated with an impact less than half that size.
  - Impact of inflation on credit did not change over the financial cycle.
- Caveats on interpreting coefficient changes:
  - Crisis dummies indicate relative importance of supply vs demand, not absolute changes in importance.
  - Asymmetry: credit (a stock variable) can adjust up more flexibly than down; declines are constrained by timing of debt repayments.

### Foreign ownership: effects and mechanisms
- Summary of findings:
  - Foreign ownership is associated with significantly higher credit growth after controlling for other factors (foreign dummy highly significant in column (1) of Table 1).
  - The effect of foreign ownership, when interacted with different crisis years (column (2) of Table 1), diminished over time.
- Differences between foreign and domestic banks (column (3) of Table 1):
  - Domestic banks grew faster when more profitable (higher return on equity); for foreign banks this appears true only during the crisis period.
    - Possible reason: prior to the crisis, foreign banks did not depend as much on retained earnings to build capital and grow.
  - Foreign banks’ credit expansion was more sensitive to solvency indicators (equity to net loan ratio) than domestic banks’, throughout the cycle.
    - Suggests foreign banks had greater propensity to leverage and deleverage: used additional capital to leverage up more than domestic banks; loss of bank equity affected credit growth more in foreign banks.
  - Foreign banks reacted more negatively than domestic banks to lower asset quality (higher reserves to gross loans coefficient) throughout the cycle.
  - Foreign banks did not respond differently from domestic ones to global and domestic macro factors.
- Parent bank characteristics among foreign banks:
  - Ownership effect on subsidiary credit growth depends on parent home country CDS spreads and parent solvency.
  - Stronger sovereign (lower CDS) helps due to lower funding costs transmitted to the parent and greater likelihood of contingent sovereign support.
  - The effect of parent solvency on subsidiary credit growth became much weaker after 2008, reflecting parents’ greater need to accumulate capital and greater autonomy given to subsidiaries since the crisis.
- Additional note:
  - Average equity to net loan ratios for domestic banks were higher than for foreign banks in both pre- and post-crisis periods, suggesting foreign banks used more leverage.

### Key regression results (Table 1: CESEE: Determinants of Credit Growth (2001–11))
- Dependent variable: real annual loan growth (in percent)
- Coefficients and standard errors (selected):
  - EMBIG spread:
    - Column (1): -0.014*** (0.005)
    - Column (2): -0.019*** (0.005)
    - Column (3): -0.024*** (0.006)
  - Real domestic demand growth:
    - Column (1): 0.977*** (0.100)
    - Column (2): 1.794*** (0.199)
    - Column (3): 1.544*** (0.194)
    - × crisis dummy (Column (2)): -1.252*** (0.235)
    - × crisis dummy (Column (3)): -0.865*** (0.222)
  - Average inflation:
    - Column (1): -0.806*** (0.119)
    - Column (2): -0.754*** (0.119)
    - Column (3): -0.656*** (0.117)
  - Bank size to GDP (first lag):
    - Column (1): -2.911*** (0.428)
    - Column (2): -3.804*** (0.530)
    - Column (3): -3.146*** (0.498)
    - × crisis dummy (Column (2)): 0.852*** (0.263)
  - Reserves to gross loans ratio (first lag):
    - Column (1): -1.238*** (0.209)
    - Column (2): -0.629*** (0.219)
    - × crisis dummy (Column (2)): -0.763** (0.346)
    - × crisis dummy (Column (3)): -1.077*** (0.403)
    - × foreign dummy (Column (3)): -1.785*** (0.440)
    - × crisis × foreign dummy (Column (3)): 1.373** (0.534)
  - Net loan to deposit ratio (first lag):
    - Column (1): -1.238*** (0.209)
    - Column (2): -0.062*** (0.018)
    - × crisis dummy (Column (2)): -0.066*** (0.018)
  - Equity to net loan ratio (first lag):
    - Column (1): 0.497*** (0.059)
    - Column (2): 0.428*** (0.060)
    - Column (3): 0.402*** (0.066)
    - × crisis dummy (Column (2)): 0.626*** (0.106)
    - × crisis dummy (Column (3)): 0.309*** (0.087)
    - × foreign dummy (Column (3)): 0.643*** (0.119)
  - Liquid assets to deposits and st funding ratio (first lag):
    - Column (1): 0.212*** (0.053)
    - Column (2): 0.258*** (0.059)
    - Column (3): 0.219*** (0.054)
    - × crisis dummy (Column (2)): -0.222*** (0.078)
  - Return on equity (first lag):
    - Column (1): 0.207*** (0.058)
    - Column (2): 0.163* (0.085)
    - × crisis dummy (Column (2)): 0.165** (0.074)
    - × foreign dummy (Column (3)): -0.410*** (0.157)
    - × crisis × foreign dummy (Column (3)): 0.587*** (0.149)
  - Foreign dummy:
    - Column (1): 19.511*** (3.648)
    - Column (2): 19.965*** (3.578)
  - Parent bank equity to net loan ratio (first lag) (Column (3)): 2.479*** (0.435)
  - Parent bank home CDS spread (Column (3)): -0.037*** (0.010)
  - Crisis year dummies (selected, Column (1)/(3) formatting in table):
    - dum_crisis: -4.808* (2.543)
    - dum_foreign_crisis: -5.092* (2.920)
    - dum_f2008: 7.745** (3.717)     × crisis dummy: -1.951*** (0.442)
    - dum_f2009: -5.672 (3.768)
    - dum_f2010: -6.747* (3.478)
    - dum_f2011: -8.236** (3.560)
- Sample and fit statistics:
  - Number of observations:
    - Columns (1) and (2): 2,415
    - Column (3): 2,093
  - R-squared:
    - Column (1): 0.336
    - Column (2): 0.372
    - Column (3): 0.404
  - Number of banks:
    - Columns (1) and (2): 435
    - Column (3): 415
  - Bank fixed effects: yes (all reported specifications)
- Sources and estimation:
  - Sources: Bankscope, Bloomberg, WEO database, and authors' calculations
  - Results obtained through a fixed effect estimation. Standard errors reported in parentheses. ***, **, and * indicate a p-value lower than 1 percent, 5 percent, and 10 percent respectively.
  - A dummy for Belarus in 2010 is included to account for a break in the series for Belarusian banks in that year.

### Transition to case studies (preview)
- Section III explores credit demand and supply in five case studies: Latvia, Lithuania, Montenegro, Poland, and Romania.
- Methodology for case studies:
  - Joint estimation of credit demand and credit supply using a disequilibrium model where actual (new) lending equals the lower of estimated demand or supply of credit.
  - The model assesses whether credit demand or supply constrained actual credit evolution and allows for quantity rationing beyond price (interest rate) adjustments.
  - The flow of new lending (in real terms) is used (different from real credit growth used in panel regressions).
  - Appendix III contains details.
- Empirical context (Figure 3 highlights):
  - After the boom, GDP contracted sharply (except in Poland).
  - Credit growth reversed and was negative for some countries for several years.
  - Credit-to-GDP adjusted heterogeneously across countries.
  - Parent funding reversals varied by country.
  - Loan-to-deposit ratios have generally come down (except Romania).
  - NPLs increased, with sectoral heterogeneity.

*Source: _wp1515 - Box 2. Estimation and Model Specification of Panel Regression (continued)*

### Box 3. Estimation and Model Specification of Disequilibrium Model

### Box 3. Estimation and Model Specification of Disequilibrium Model

### Model structure and assumptions
- Credit supply and credit demand are simultaneously estimated in a system of equations with endogenous switching proposed by Laffont and Garcia (1977).
- Underlying assumption: the interest rate is not perfectly flexible to clear the market; non-price factors also determine supply and demand for credit, allowing the market to be in disequilibrium and to exhibit quantity rationing.
- Observed new lending Ct is assumed to be the minimum of the estimated demand for credit and estimated supply for credit:
  - Observed new lending = min(estimated demand, estimated supply)
- The vectors of explanatory variables for demand and supply contain the interest rate and non-price determinants. Error terms are assumed jointly normal and independent over time, with a zero mean and covariance matrix ∑.

### Estimation technique and robustness
- Estimation is performed using the maximum likelihood method proposed by Maddala and Nelson (1974).
- Switching regression technique implies greater sensitivity to specification because the observed dependent variable equals only one of the dependent variables in the model, with the other unobservable.
- Insignificant variables are usually dropped to improve stability and precision, except the interest rate which is always retained in both equations.
- Model fit is assessed by comparing actual credit with the minimum of either demand or supply (see Appendix IV referenced in source).
- Statistical significance of estimated excess supply is assessed using Monte Carlo simulations to compute confidence intervals.
  - For significance bands, the point estimate of the parameter vector, the estimated variance covariance matrix, and normally distributed shocks are used to generate alternative parameter vectors.
  - Using the alternative parameter vector and observed regressors (same across repetitions), fitted values for demand, supply and excess supply are constructed.
  - This process is repeated a large number of times (typically 5,000).
  - For each observation, the mean and the standard deviation of the predicted excess supply across repetitions are computed.
- Fixed-error confidence bands are reported (derived from averaging the standard deviation of the predicted excess supply across observations). Time varying-confidence bands are provided as a robustness check in Appendix V.

### Data treatment and dependent variable
- The dependent variable is new loans extended (a flow variable), in real terms.
- Real new loans include both new loan contracts and rollovers of existing loans.
- Either monthly flows, or a 3- or 6-month smoothed average of monthly flows are used.
  - Moderate smoothing is used where monthly data noise interferes with estimation; smoothing may introduce potential endogeneity concerns (e.g., credit at time t-2 impacting indicators at time t), but averages out timing between decision to obtain financing and time credit was obtained.
- In three of the five country cases, regressions are estimated separately for new loans to households and new loans to non-financial corporates (NFCs), depending on data availability.
- Estimation uses monthly data with interpolations of quarterly data where needed.

### Choice of explanatory variables (a priori exclusion restrictions and expected signs)
- Lending rate enters both demand and supply equations.
- Demand-side explanatory variables (expected sign indicated):
  - The cost of credit: the real lending rate (-), inflation expectations (+).
  - Economic conditions: confidence surveys (+), changes in stock market indices (+), indicators of current economic activity such as real GDP, retail sales, industrial production, or new orders (+), uncertainty about the future proxied by the volatility of consensus forecasts (-); for corporates and households: profitability prospects based on survey data (+), real wage growth (+), employment growth (+), unemployment rate (-).
  - Debt overhang: debt stocks in percent of GDP (-), corporate and household NPL ratio (-).
  - Alternative funding sources for corporate borrowers: profit or cash developments (-), stock market returns (-), surveys on firms’ financial constraints (+).
- Supply-side explanatory variables (expected sign indicated):
  - The return on credit: the real lending rate (+), interest margin (+), inflation expectations (-).
  - Economic conditions: confidence surveys (+), changes in stock market indices (+), indicators of current economic activity such as real GDP growth (+), value of collateral (e.g., real estate prices (+)).
  - Debt overhang and borrower creditworthiness: NPL ratio (-).
  - Funding costs for banks or indicators of financial stress: real deposit or other funding rate (-).
  - Capacity to lend: deposits and/or parent funding (+), banking system capital divided by minimum capital requirements (+).

### Treatment of parent funding and endogeneity
- Parent funding enters the supply-side equation; a two-step estimation procedure is used to control for potential endogeneity.
  - Concern: parent funding increases capacity to lend (supply) but parent banks can also expand/contract funding to subsidiaries in response to credit demand.
  - Two-stage procedure:
    - First stage: regress parent funding (or change therein) on instruments using ordinary least squares. Instruments used are parent bank (or home country) CDS spreads, which capture health/stress of parent bank balance sheets.
    - Second stage: fitted values from first stage are included in the supply equation. The fitted values reflect part of parent funding attributable to parent bank health or stress.
  - The fitted parent-funding variable captures a narrower concept of variability in parent funding (only that part related to parent balance sheet health/stress). The significance of this variable must be narrowly interpreted.
- For some countries, parent funding is not strongly related to parent bank stress (first stage), or local lending is not driven by parent funding that responds to parent bank stress; specification is decided pragmatically for each country.

### Model performance, sensitivity, and general findings across cases
- Model shows sensitivity to estimation specification, but both demand and supply are generally estimated within reasonable error bands.
- Predicted credit and actual credit are fairly close (figures in Appendix IV).
- Most coefficients have expected signs, but the interest rate is sometimes not significant.
- Estimation results are somewhat sensitive due to:
  - Estimating the path of two unobservable variables over time.
  - Limitations of maximum likelihood estimation that could converge on a local rather than a global optimum.
- Across case studies, both credit demand and supply factors matter and vary over time:
  - Both demand and supply rose in tandem during the boom and jointly fell in the bust for all countries studied.
  - Timing and extent of when credit demand exceeds credit supply (and vice versa) vary by country.
  - Examples:
    - Montenegro and Lithuania: credit supply became more constraining in the post-crisis period (consistent with panel regressions in Section II).
    - Latvia: simultaneously tightening supply and demand contributed broadly equally to the contraction (neither was overriding), with some role for demand constraints by NFCs.
    - Poland and Romania: demand factors constrained credit most of the time.

### Montenegro — key findings and chronology
- Supply constraints matter most in explaining credit expansion throughout the sample for Montenegro.
- Model finds several periods of statistically significant excess demand (i.e., supply constraints) during the boom, but no periods of excess supply.
- Other periods show disequilibria not statistically significant, implying both demand and supply played an equal role.
- Evolution described in three periods:
  - Post-independence (before 2008):
    - Period shortly after independence attracted large capital inflows, especially in the nontradable sector.
    - Pent-up demand coexisted with insufficient credit supply even as supply rose rapidly.
    - Inflows of parent bank funding increased from 7 to 27 percent of GDP between 2006 and 2008.
    - Excess demand quickly disappeared by mid-2007.
  - Crisis (2008–10):
    - Model estimates credit supply leveled off as early as 2008 while credit demand continued to rise until mid-2008; by mid-2008, credit demand starts to fall and excess demand disappears by end-2008.
    - Leveling off of estimated supply occurs somewhat before massive deposit withdrawals (from mid-2008 onward) and large reversals of parent bank funding (from mid-2009 onward).
    - NPLs were already rising in 2008, exceeding 7 percent by the end of that year (much above the average of 4 percent of the other four countries in the case study).
  - Economic recovery (2011–12):
    - Recovery associated with rising credit demand, but supply either continued to contract or stabilized (did not rise), resulting in supply constraints.
- Box 4: Estimation performance for Montenegro
  - Joint estimation of credit to NFCs and households (data could not be obtained separately).
  - Demand side: interest rate has expected negative sign. Credit demand positively correlated with real economic activity and negatively with proxy for debt overhang.
  - Supply side: interest margin insignificant; deposit growth significantly and positively associated with credit expansion. NPL ratio negatively associated with credit supply (consistent across countries).
  - Parent funding (instrumented by sovereign CDS spreads of home countries) is significant and positive in the supply equation; coefficient lower than on deposit funding.
- Box 5: Montenegro contextual statistics and developments
  - GDP per capita (US$ PPP, 2012): 11,800
  - Population (2012): 622,000
  - Exchange rate regime: uses Euro
  - During 2004–08, GDP growth averaged 7 percent.
  - Credit-to-GDP ratio increased from 38 to 89 percent of GDP between 2006 and 2008.
  - Parent bank funding increased from 7 to 27 percent of GDP between 2006 and 2008.
  - Loan-to-deposit ratio peaked in 2008 at 141 percent.
  - Parent bank funding fell from 27 to just over 10 percent of GDP between 2008 and 2012.
  - Credit growth was negative for four consecutive years (2009–12).
  - Credit-to-GDP ratio fell by some 34 percentage points of GDP over 2009–12 (largest decline among the case studies).
  - NPLs rose and remained high, creating challenges for resolution and restructuring.
  - Central bank measures included broadening and tightening reserve requirement calculations in 2006 and 2007–08, increasing capital adequacy ratio from 8 to 10 or 12 percent for banks with credit growth in excess of 60 or 100 percent respectively, and capping annual credit growth at 30, 40, or 60 percent for banks with outstanding loans above 200 million, between 100 and 200 million, and below 100 million euro respectively.
  - Central bank later reduced effective reserve requirement rates and reduced the interest rate for reserve requirements for liquidity.

*Source: IMF staff estimates and discussion in Box 3: "Estimation and Model Specification of Disequilibrium Model."*

### Box 6. Lithuania. Estimation Performance

### Box 6. Lithuania. Estimation Performance

### Joint estimation approach
- The model for Lithuania is estimated for new credit to the nonfinancial private sector (with credit for households and NFCs jointly), since credit to each sector evolved at about the same pace in Lithuania and therefore the two series are highly correlated.

### Significance and interpretation of coefficients
- Most coefficients have the expected sign and are significant.
- Demand equation:
  - Profit margins are significantly and positively related to credit demand.
- Supply equation:
  - The NPL ratio has a strong and significant role in determining supply of credit to private sector, with a higher NPL ratio being correlated with lower credit supply.

### Parent funding and funding channels
- In the supply equation, the size of the coefficient of parent funding, instrumented by the CDS of parent banks, is much larger than the coefficient on deposits, indicating a greater responsiveness of credit supply to parent funding than to deposit funding.

### Dynamics of demand and supply since the bust
- Demand and supply for credit fell strongly after the 2008/09 crisis; the economic recovery has not been matched by a recovery in credit demand or credit supply.
- The model estimates two short periods of small, but statistically significant imbalances:
  - A short period of excess supply in 2009Q2–Q3, which appears to result from the slow adjustment of credit supply to the crisis.
  - Credit market imbalances become statistically significant from early 2012 onward, with credit demand somewhat exceeding credit supply—consistent with the findings of the panel regressions.
- The period of early 2012 onward coincides with a period of significant excess liquidity in the banking system.
- Despite excess liquidity, factors such as difficult NPL resolution and strict lending standards may explain why supply constraints dominate during this period; anecdotal evidence confirms that NPL resolution has been particularly slow in Lithuania.

### Relevant contextual notes (from accompanying boxes/footnotes)
- Excess supply in 2009Q2–Q3: "At this time credit demand is estimated to be sharply contracting, but deleveraging has not yet geared full speed. Excess supply disappears as soon as the first deleveraging episode starts (see Box 7 for a description of deleveraging episodes)."
- Excess liquidity and depositor payout: "This period follows the Snoras depositor payout (4 percent of GDP), which led to an increase in liquidity in foreign owned banks. This, in turn, signals the start of the second deleveraging episode in Lithuania (see Box 7). This means that banks at that time were not constrained by capacity to lend."

_ Sources: Central Bank of Lithuania; Haver; and IMF staff estimates. _

### Box 10. Poland. Estimation Performance

### Box 10. Poland. Estimation Performance

### Model specification
- The model for Poland is estimated separately for new credit to households and new credit to NFCs.
- Household credit in Poland expanded particularly strongly, and mortgages in foreign exchange were prevalent (see Box 11).

### Statistical significance and coefficients
- All variables are significant, except:
  - lending rate in the supply equation for households; and
  - lending rates and deposit rates for NFCs.
- The NPL ratio is significant in both the household and NFC supply equations.
- The coefficient on debt overhang variables is larger in the case of credit to households compared to credit to corporates, implying higher sensitivity of banks towards impaired household credit developments.

### Parent funding
- The coefficient of parent funding, instrumented by CDS spread of parent banks, is significant with the predicted sign in the supply equation for credit to household.
- The R square of the first-step OLS regression is very low.

*Source: IMF staff estimates (Box 10, Poland, Estimation Performance).*

### REFERENCES

### _wp1515 - REFERENCES

### References (selected)
- Amiti, Mary and David E. Weinstein, 2013, “How Much do Bank Shocks Affect Investment? Evidence from Matched Bank-Firm Loan Data,” NBER Working Paper No. 18890.
- Aydin, Burcu, 2008, “Banking Structure and Credit Growth in Central and Eastern European Countries,” IMF Working Paper WP/08/215 (Washington, D.C.; International Monetary Fund).
- Bakker, Bas B. and Anne-Marie Gulde, 2010, “The Credit Boom in the EU New Member States: Bad Luck or Bad Policies?,” IMF Working Paper WP/10/130 (Washington, D.C.; International Monetary Fund).
- Barajas, Adolfo and Roberto Steiner, 2002, “Credit Stagnation in Latin America”, IMF Working Paper WP/02/53 (Washington, D.C.; International Monetary Fund).
- Berger, Allen N. and Gregory F. Udell, 1992, “Some Evidence on the Empirical Significance of Credit Rationing,” Journal of Political Economy, Vol. 100, pp.1047-1077.
- Bonin, John, Iftekhar Hasan, and Paul Wachtel, 2005, “Bank Performance, Efficiency, and Ownership in Transition Countries,” Journal of Banking and Finance, Vol. 29, pp 31-53.
- Brown, Martin, Steven Ongena, Alexander Popov, and Pinar Yesin, 2012, ”Who Needs Credit and Who Gets Credit in Eastern Europe?,” ECB Working Paper No. 1421.
- Calvo, Guillermo A., Alejandro Izquierdo, and Ernesto Talvi, 2006, “Phoenix Miracles in Emerging Markets: Recovering without Credit from Systemic Financial Crises”, NBER Working Paper No. 12101.
- Claessens, Stijn, Asli Demirgüç-Kunt and Harry Huizinga, 2001, “How Does Foreign Entry Affect Domestic Banking Markets?,” Journal of Banking and Finance, Vol. 25, pp891-911.
- Cottarelli, Carlo, Giovanni Dell’Ariccia, and Ivanna Vladkova-Hollar, 2003, “Early Birds, Late Risers, and Sleeping Beauties: Bank Credit Growth to the Private Sector in Central and Eastern Europe and the Balkans,” IMF Working Paper WP/03/213 (Washington, D.C.; International Monetary Fund).
- Degryse, Hans, Olena Havrylchyk, Emilia Jurzyk, and Sylwester Kozak, 2009, “Foreign Bank Entry and Credit Allocation in Emerging Markets,” IMF Working Paper WP/09/270 (Washington, D.C.; International Monetary Fund).
- Demirgüç-Kunt, Asli, and Maria Soledad Martínez Pería, 2010, “A Framework for Analyzing Competition in the Banking Sector: An Application to the Case of Jordan,” World Bank Policy Research Paper 5499.
- De Haas, Ralph and Iman Van Lelyveld, 2006, “Foreign Banks and Credit Stability in Central and Eastern Europe: A Panel Data Analysis,” Journal of Banking and Finance, Vol. 30, pp. 1927-1952.
- Duenwald, Christoph, Nikolay Gueorguiev, and Andrea Schaechter, 2005, “Too Much of a Good Thing? Credit Booms in Transition Economies: The Case of Bulgaria, Romania, and Ukraine, IMF Working Paper WP/05/128 (Washington, D.C.; International Monetary Fund).
- Égert, Balázs, Peter Backé, and Tina Zumer, 2006, “Credit Growth in Central and Eastern Europe: New (Over)Shooting Stars?,” ECB Working Paper No. 687.
- Eichengreen, Barry and Katharina Steiner, 2008, “Is Poland at Risk of a Boom-And-Bust Cycle in the Run-Up to Euro Adoption?,” NBER Working Paper No. 14438.
- Enoch, Charles and Inci Ötker-Robe, 2007, “Rapid Credit Growth in Central and Eastern Europe: Endless Boom or Early Warning?,” IMF, Palgrave Macmillan, April.
- European Banking Coordination “Vienna” Initiative, Working Group on NPLs in Central, Eastern and Southeastern Europe, March 2012. www.imf.org/external/region/eur/pdf/2012/030112.pdf
- Fritsch, Markus, Fabian Gleisner, and Markus Holzhäuser, 2007, “Bank M&A in Central and Eastern Europe,” mimeo, Johann Wolfgang Goethe-University and E-Finance Lab mimeo, February 2007, Available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=966294
- Gabaix, Xavier, 2011, “The Granular Origins of Aggregate Fluctuations,” Econometrica, Vol. 79 (May), No. 3, pp 733-772.
- Ghosh, Atish R. and Swati. R. Ghosh, 1999, “East Asia in the Aftermath: Was There a Crunch?,” IMF Working Paper WP/99/38 (Washington, D.C.; International Monetary Fund).
- Gorton, Gary and Ping He, 2005, “Bank Credit Cycles,” NBER Working Paper No. 11363.
- Havrylchyk, Olena and Emilia Jurzyk, 2010, “Inherited or Earned? Performance of Foreign Banks in Central and Eastern Europe,” IMF Working Paper WP/10/4 (Washington, D.C.; International Monetary Fund).
- Hilbers, Paul, Inci Otker-Robe, Ceyla Pazarbasioglu, and Gudrun Johnsen, 2005, “Assessing and Managing Rapid Credit Growth and the Role of Supervisory and Prudential Policies,” IMF Working Paper WP/05/151 (Washington, D.C.; International Monetary Fund).
- Impavido, Gregorio., Jérôme Vandenbussche, and Li Zeng, 2015, “Bank Funding Models and Credit Growth in Central, Eastern and South-Eastern Europe: Digging into the Effect of Foreign Ownership”, forthcoming IMF working Paper.
- International Monetary Fund, 2011, “Navigating Stormy Waters,” Regional Economic Outlook: Europe. October.
- International Monetary Fund, 2012, “How Emerging Europe Came through the 2008/09 Crisis: An account by the Staff of the IMF’s European Department”, edited by B. B. Bakker and C. Klingen (Washington, D.C.; International Monetary Fund).
- International Monetary Fund, 2013a, “Financing Future Growth: the Evolving Role of Banking System in CESEE”, Regional Economic Issues, April 2013 (Washington, D. C.; International Monetary Fund).
- International Monetary Fund, 2013b, German-Central European Supply Chain-Cluster Report: Staff Report, First Background Note, Second Background Note, Third Background Note, (Washington, D.C.; International Monetary Fund); http://www.imf.org/external/publications/
- International Monetary Fund, 2013c, Republic of Lithuania: 2013 Article IV Consultation - Staff Report, Information Annex, Public Information Notice, Statement by the Executive Director, IMF Country Report No.13/81 (Washington, D.C.; International Monetary Fund).
- International Monetary Fund, 2014, Baltic Cluster Report: Selected Issues, IMF Country Report No.14/117 (Washington, D.C.; International Monetary Fund).
- Kapan, Tumer and Camelia Minoiu, 2013, “Balance Sheet Strength and Bank Lending During the Global Financial Crisis, IMF Working Paper WP/13/102 (Washington, D.C.; International Monetary Fund).
- Khwaja, Asim I. and Atif Mian, 2008, “Tracing the Impact of Bank Liquidity Shocks: Evidence from an Emerging Market,” American Economic Review, Vol. 98, pp. 1413-1442.
- Kim, Hyun E., 1999, “Was Credit Channel a Key Monetary Transmission Mechanism Following the Recent Financial Crisis in the Republic of Korea?”, World Bank Policy Research Working Paper No. 3003.
- Laffont, Jean-Jacques, and René Garcia, 1977, "Disequilibrium Econometrics for Business Loans," Econometrica, Vol. 45 (July), pp. 1187-120.
- Lane, Philip R. and Peter McQuade, 2014, “Domestic Credit Growth and International Capital Flows,” Scandinavian Journal of Economics, Vol. 116, pp. 218-252.
- Maddala, Ganggadharrao S. and Forrest. D. Nelson, 1974, “Maximum Likelihood Methods for Models of Market Disequilibrium”, Econometrica, 42(6): 1013-30.
- Mendoza, Enrique and Marco Terrones, 2012, “An Anatomy of Credit Booms and Their Demise,” NBER Working Paper No. 18379.
- Nier, Erlend and Lea Zichhino, 2008, “Bank Losses, Monetary Policy, and Financial Stability—Evidence on the Interplay from Panel Data,” IMF Working Paper WP/08/232; (Washington, D.C.; International Monetary Fund).
- Pazarbaşioğlu, Ceyla, 1997, “A Credit Crunch? Finland in the Aftermath of the Banking Crisis”, IMF Staff Papers, 44(3): 315-27 (Washington: International Monetary Fund).
- Poghosyan, Tigran, 2010, “Slowdown of Credit Flows in Jordan in the Wake of the Global Financial Crisis: Supply or Demand Driven?,” IMF Working Paper WP/10/256; (Washington, D.C.; International Monetary Fund).
- Rosenberg, Christoph B., and Marcel Tirpak, 2008, “Determinants of Foreign Currency Borrowing in the New Member States of the EU”, IMF Working Paper WP/08/173; (Washington, D.C.; International Monetary Fund).
- Tressel, Thierry and Enrica Detragiache, 2008, “Do Financial Sector Reforms Lead to Financial Development?,” IMF Working Paper WP/08/265 (Washington, D.C.; International Monetary Fund).
- Vandenbussche, Jérôme, Ursula Vogel, and Enrica Detragiache, 2015, “Macroprudential Policies and Housing Prices—A New Database and Empirical Evidence for Central, Eastern, and Southeastern Europe,” Journal of Money, Credit and Banking (forthcoming).
- Waters, George A., 2012, “Quantity Rationing of Credit,” Bank of Finland Research Discussion Papers No. 3.

### Appendix I. Data Used in the Panel Regression — Country and Time Distribution
- Country distribution (Number of observations by bank type, totals preserved):
  - Albania: Domestic banks 104, Foreign banks 656
  - Belarus: Domestic banks 423, Foreign banks 779
  - Bosnia & Herzegovina: Domestic banks 5280, Foreign banks 132
  - Bulgaria: Domestic banks 7176, Foreign banks 147
  - Croatia: Domestic banks 174, Foreign banks 105279
  - Czech Republic: Domestic banks 2010, Foreign banks 7127
  - Estonia: Domestic banks 1720, Foreign banks 37
  - Hungary: Domestic banks 164662
  - Latvia: Domestic banks 8663, Foreign banks 149
  - Lithuania: Domestic banks 444690
  - Macedonia, FYR: Domestic banks 464, Foreign banks 187
  - Moldova: Domestic banks 581, Foreign banks 472
  - Montenegro, Rep. of: Domestic banks 172, Foreign banks 542
  - Poland: Domestic banks 3197, Foreign banks 128
  - Romania: Domestic banks 40128, Foreign banks 168
  - Serbia, Republic of: Domestic banks 7473, Foreign banks 147
  - Slovak Republic: Domestic banks 1110, Foreign banks 1112
  - Slovenia: Domestic banks 6343, Foreign banks 106
  - Turkey: Domestic banks 6659, Foreign banks 125
  - Ukraine: Domestic banks 167103, Foreign banks 270
  - Total: Domestic banks 1105, Foreign banks 13102415
  - Sources: Bankscope and IMF staff estimates.

- Time distribution (No. of observations by year):
  - No. of obs. for domestic banks: 2001 85, 2002 90, 2003 96, 2004 102, 2005 107, 2006 105, 2007 102, 2008 109, 2009 106, 2010 96, 2011 105
  - No. of obs. for foreign banks: 2001 44, 2002 52, 2003 65, 2004 75, 2005 101, 2006 126, 2007 148, 2008 161, 2009 171, 2010 187, 2011 180, 2012 131, 2013 0, 2014 0 (table shows totals per year culminating in Total 2415)
  - Total observations by year: 129, 142, 161, 177, 208, 233, 253, 263, 280, 293, 276, 2415
  - Sources: Bankscope, and IMF staff estimates.

### Appendix I. Summary Statistics of Data (Table AI.3)
- Variable — No. of observations — Mean — Standard deviation — Min — Max
  - Growth of gross loans (%) — 2415 — 21.93 — 37.1 — -54.2 — 345.5
  - EMBIG spread (pp) — 2415 — 392.1 — 164.4 — 19.7 — 1796.4
  - Real domestic demand growth (percent) — 2415 — 4.1 — 8.1 — -27.4 — 24.2
  - Average inflation (percent) — 2415 — 6.5 — 7.3 — -1.2 — 80.6
  - Bank size (% of host country GDP, 1st lag) — 2415 — 3.8 — 5.5 — 0.0 — 43.4
  - Reserves to gross loan ratio (%, 1st lag) — 2415 — 5.6 — 5.2 — 0.0 — 41.9
  - Net loans to customer deposits ratio (%, 1st lag) — 2415 — 109.0 — 74.9 — 12.9 — 679.1
  - Liquidity to dep. & st funding ratio (%, 1st lag) — 2415 — 38.2 — 22.9 — 1.3 — 246.8
  - Equity to net loans ratio (%, 1st lag) — 2415 — 27.3 — 21.7 — 1.7 — 215.7
  - Return on average equity (%, 1st lag) — 2415 — 8.4 — 15.1 — -99.1 — 75.7
  - Parent equity to total assets ratio (%) — 988 — 6.3 — 4.9 — -85.8 — 40.5
  - Parent bank home country CDS spreads — 988 — 99.8 — 134.5 — 0.0 — 812.4
- Sources: Bankscope; and IMF staff estimates.
- Note: Summary statistics of parent bank variables are only reported for foreign bank observations included in the regressions with those variables.

### Appendix II. Case Study Summary Statistics and Narratives — Key Indicators (Tables AII.1–AII.2)
- Table AII.1. Summary Indicators (percent) — Periods and country entries (preserved as in source):
  - Periods shown: 2004-08, 2009-12 across countries Latvia, Lithuania, Montenegro, Poland, Romania.
  - Average real GDP growth:
    - Latvia 2004-08: 7.0; 2009-12: -0.1
    - Lithuania 2004-08: 7.1; 2009-12: -1.2
    - Montenegro 2004-08: 7.3; 2009-12: -1.9
    - Poland 2004-08: 5.4; 2009-12: 3.0
    - Romania 2004-08: 6.8; 2009-12: -1.9
  - Average inflation:
    - Latvia 2004-08: 4.8; 2009-12: 2.8
    - Lithuania 2004-08: 4.9; 2009-12: 3.2
    - Montenegro 2004-08: 9.0; 2009-12: 2.1
    - Poland 2004-08: 2.8; 2009-12: 3.4
    - Romania 2004-08: 8.0; 2009-12: 5.8
  - Average nominal credit growth and Average real credit growth (selected entries preserved):
    - Latvia average nominal credit growth 2009-12: -8.1; average real credit growth 2009-12: -10.9
    - Lithuania average nominal credit growth 2009-12: 42.2; average real credit growth 2009-12: 37.3
    - Montenegro average nominal credit growth 2009-12: -5.4; average real credit growth 2009-12: -8.5
    - Poland average nominal credit growth 2009-12: 41.9; average real credit growth 2009-12: 30.8
    - Romania average nominal credit growth 2009-12: -8.3; average real credit growth 2009-12: -10.1
  - Average nominal exchange rate change: entries include "peg" for many periods and numerical values for others, e.g., Montenegro 2009-11: 3.0; Poland 2009-11: 0.7; Romania 2009-11: -4.7
  - Source: Authorities and Fund staff calculations.

- Table AII.2. Banking sector indicators (end-year values) — Selected indicators across Latvia, Lithuania, Montenegro, Poland, Romania (preserving fragmented table format and values):
  - Size of banking sector (% of GDP) examples:
    - Latvia 2006.., 2008 63.0, 2012 62.7
    - Lithuania 2006 38.9, 2008 81.9, 2012 74.4
    - Montenegro 2006 89.7, 2008 144.5, 2012 130.3
    - Poland 2006 58.4, 2008 82.2, 2012 85.8
    - Romania 2006 ..72.7, 2008 68.9
  - Credit stock (percent of GDP) examples:
    - Latvia 2006 38.5, 2008 89.3, 2012 56.3
    - Lithuania 2006 21.0, 2008 60.9, 2012 48.3
    - Montenegro 2006 47.1, 2008 103.1, 2012 75.6
    - Poland 2006 26.6, 2008 47.2, 2012 50.1
    - Romania 2006 17.7, 2008 41.8, 2012 15.9
  - Mortgages (share of credit) and household lending (examples):
    - Mortgage lending shares include values such as 5.5, 18.7, 19.1, 7.6, 31.4 depending on country-year entries.
    - Lending to households in total shows entries like 27.7, 42.5, 46.5, 25.3, 39.6 across countries and years.
  - Deposit stock (% of GDP) examples:
    - Latvia 2006 38.8, 2008 47.0, 2012 46.2
    - Lithuania 2006 25.3, 2008 34.6, 2012 34.5
    - Montenegro 2006 58.6, 2008 80.4, 2012 36.6
    - Poland 2006 38.6, 2008 44.8, 2012 50.6
    - Romania 2006 ..28.8, 2008 32.4
  - Parent loans (Gross external debt of credit institutions for Poland and Romania and Montenegro) examples:
    - Latvia 2006 7.3, 2008 26.9, 2012 10.6
    - Lithuania 2006 7.3, 2008 35.4, 2012 18.1
    - Montenegro 2006 9.0, 2008 44.9, 2012 23.9
    - Poland 2006 5.6, 2008 16.5, 2012 13.7
    - Romania 2006 1.2, 2008 17.8, 2012 15.6
  - Structure of banking sector (share, in percent) examples:
    - Forex deposits in total deposits: entries include 30.1, 30.2, 31.6, 69.6, 69.4, 76.2 depending on country-year.
    - Forex lending in total lending: entries include 47.9, 62.5, 68.9, 60.6, 89.5, 88.0 depending on country-year.
    - Concentration (share of largest 5 or 4) by assets examples: 87.0, 57.0, 75.0, 63.1, 69.5, 62.8, 52.3, 44.6, 44.9, 63.9, 54.3, 55.2
  - Degree of foreign ownership (by assets and by loans) — selected entries:
    - By assets of foreign-owned banks: examples include 58.8, 71.7, 74.0, 86.0, 62.6, 62.5, 67.8, 72.3, 63.6, 82.2, 81.2
    - By loans of foreign-owned banks: examples include 64.0, 73.1, 74.5, 87.5, 70.7, 75.9, 71.7, 0.0
  - Other indicators (percent) — selected end-year values:
    - Real lending rate examples: 7.8, 0.4, 5.9, -5.2, 0.1, 0.0, 1.3, 1.8, 7.3, 5.3, 11.3, 8.7, 8.3
    - NPL ratio examples: 7.2, 16.9, 4.6, 13.9, 3.6, 11.1, 21.2, 4.4, 8.9, 2.8, 18.2
    - Loan-to-deposit ratio examples: 78.9, 140.6, 98.5, 187, 121, 143.2, 273.4, 174.9, 77.9, 120.0, 110.2, 77.2, 142.2, 154.5
  - Notes from table:
    - 1/ For Lithuania, share of largest 4.
    - 2/ Gross external debt of credit institutions for Poland and Romania and Montenegro.
    - 3/ Not from related MFIs, i.e. excluding parent loans.
    - 4/ For Romania, 2004 data.
    - 5/ 2004 data for Lithuania.
    - 6/ 2009 data instead of 2008 data for Latvia.

*Source: _wp1515 - REFERENCES (PDF).*

### Appendix III. Description of Variables of Disequilibrium Model

### Appendix III. Description of Variables of Disequilibrium Model

### Dependent variable (new credit flow)
- Montenegro
  - new credit flow in real terms
  - monthly data
  - in logs
  - flow data constructed from monetary survey data using the maturity structure of existing credit
  - six month moving average
  - deflated using HICP
- Lithuania
  - new credit flow in real terms
  - monthly data
  - in logs
  - flow data from BoL
  - six-month moving average of seasonally adjusted series to smooth data
  - deflated using HICP
- Latvia
  - new credit flow in real terms
  - monthly data
  - in logs
  - flow data constructed from monetary survey data using the maturity structure of existing credit; adjusted for write-offs and the removal of liquidated banks from the statistics
  - deflated using HICP
- Poland
  - new credit flow in real and foreign exchange adjusted terms
  - monthly data
  - in logs
  - flow data constructed from monetary survey data using assumptions of the amortization based on the original maturity of credit stock
  - computed at constant exchange rates, assuming that all foreign currency denominated household credit is in Swiss franc, and that all foreign currency denominated corporate credit is in euro
  - 3-month moving average of seasonally adjusted series to smooth data
  - deflated using HICP
- Romania
  - new credit flow in real terms
  - monthly data
  - in logs
  - flow data constructed from monetary survey data including both local currency and foreign exchange credit
  - flow data are computed at constant exchange rates, using the maturity structure of existing credit
  - deflated using HICP

### Credit demand equation — From (borrower type)
- Montenegro: From households and NFC
- Lithuania: From households and NFC
- Latvia: From NFC
- Poland: From NFC
- Romania: From NFC

### Cost of credit (demand-side findings)
- Montenegro
  - average lending rate, real terms, deflated by HICP
  - negative and significant
- Lithuania
  - lending rate on new loans, constructed as the weighted average of loans in litas and euro, real terms, deflated by HICP
  - negative and significant
- Latvia
  - Lending rate on new loans to non-financial corporates (in percent)
  - Constructed as the weighted average of loans in lats and in euros (due to data availability, data were used for lending up to 0.25 million euro and up to 1 year; correlation with other lending rates that are available at lower frequency was very high)
  - real terms, deflated by HICP
  - positive and insignificant
- Poland
  - lending rate for new zloty corporate loans, in real terms, deflated by HICP
  - positive and insignificant
- Romania
  - lending rate to corporate, weighted by currency of denomination of credit, nominal terms, average lending rate
  - negative and significant
  - inflation rate, percent change in CPI compared with previous month; (inflation is very volatile in Romania, and the model performed poorly using the real lending rate; hence, inflation is included separately)
  - positive and significant

### Economic conditions (demand-side indicators and significance)
- Montenegro
  - tourist arrivals, y-o-y change, six months average
  - positive and significant
  - construction new orders, six months average
  - positive and significant
  - retail sales, six months average
  - positive and significant
- Lithuania
  - economic confidence indicator, one month lagged
  - positive and insignificant
  - real change in Vilnius stock exchange index, one month lagged
  - positive and significant
  - weighted average of volatility of consensus forecasts for Lithuania’s major trading partners, to proxy for uncertainty, lagged one month
  - negative and significant
- Latvia
  - economic sentiment indicator (index, seasonally adjusted, Eurostat)
  - positive and significant
  - industry new orders, lagged by one month, seasonally adjusted series
  - positive and significant
  - unemployment, lagged one month, seasonally adjusted series
  - negative and significant
- Poland
  - industrial production, in real terms, deflated by HICP, seasonally adjusted series, one month lagged
  - positive and significant
- Romania
  - linearly interpolated seasonally adjusted real GDP growth, as a proxy for expectations
  - positive and significant
  - other indicators such as industrial production, employment, and various survey indicators proved a poor fit to the model and were dropped

### Debt overhang (demand-side)
- Montenegro
  - credit to GDP ratio, change in six months
  - negative and significant
- Lithuania
  - NPL ratio for corporate loans, monthly data since September 2008, quarterly data were used before that and interpolated
  - negative and significant
- Latvia
  - NPL ratio on corporate lending, one month lagged
  - negative and significant
- Poland
  - share of debts past due in total loans
  - negative and significant
- Romania
  - NPL ratio for corporate loans, one month lagged
  - negative and significant

### Alternative funding / other demand-side factors
- Montenegro
  - business survey indicator on whether firms are financially constraint, lagged one month
  - positive and significant
- Lithuania
  - Profitability index in the private sector (all sectors excluding public administration, education and social work); constructed as real labor productivity divided by real wages; quarterly data were interpolated to obtain monthly frequencies
  - negative and significant
- Latvia
  - difference between productivity growth and real wage growth in the private sector, as a proxy for growth in return on capital or profitability, one month lagged
  - negative and significant
- Poland
  - productivity growth (the ratio of real GDP to employment)
  - negative and significant
- Romania
  - gross profit margin of firms, lagged one month profitability to measure prospects for business expansion
  - positive and significant

### Household demand — cost of credit, economic conditions, debt overhang (household-specific)
- Cost of credit
  - Montenegro: Lending rate on new loans to households; constructed as the weighted average of housing and other loans in lats and in euros, in real terms, deflated by HICP — positive and significant
  - Lithuania: lending rate on new zloty household credit, in real terms — negative and significant
  - Latvia: Swiss Franc Libor rate, in real terms — positive and significant
  - Poland: average lending rate to households weighted by currency, nominal terms (inflation was insignificant and hence dropped) — negative and significant
- Economic conditions (households)
  - Montenegro: Percentage change (mom) in compensation of employees; seasonally adjusted series, in real terms (deflated by HICP); quarterly data were interpolated to obtain monthly frequencies — significant and positive
  - Lithuania: real change in Warsaw stock exchange index, one month lagged — positive and significant
  - Latvia: unemployment rate, one month lagged, in seasonally adjusted terms — negative and significant
  - Poland: linearly interpolated seasonally adjusted real GDP growth — positive and significant; other proxies were dropped
- Debt overhang (households)
  - Montenegro: NPL ratio for household lending; monthly data since September 2008, quarterly data before that were interpolated to obtain monthly frequencies — negative and significant
  - Lithuania: share of debts past due in total lending — negative and significant

### Credit supply equation — For households and NFC
- Return on credit (supply-side)
  - Montenegro
    - average lending rate, real terms, deflated by HICP
    - positive and significant
  - Lithuania
    - lending rate on new loans, constructed as the weighted average of loans in litas and euro, real terms, deflated by HICP
    - positive and significant
  - Latvia
    - Lending rate on new loans to non-financial corporate; constructed as the weighted average of loans in lats and in euros (due to data availability, data were used for lending up to 0.25 million euro and up to 1 year, correlation with other lending rates that are available at lower frequency was very high), real terms, deflated by HICP
    - positive and significant
  - Poland
    - lending rate for new zloty loans, in real terms, deflated by HICP
    - positive and insignificant
  - Romania
    - lending rate, average rate, nominal terms
    - positive and significant
    - inflation, percent change in CPI compared with previous month; inflation is included separately for Romania and is negative and significant

### Economic conditions (supply-side)
- Montenegro
  - economic confidence indicator, one month lagged
  - positive and significant
  - real change in Vilnius stock exchange index, one month lagged
  - positive and significant
- Latvia
  - stock exchange index
  - positive and significant
- Poland
  - linearly interpolated seasonally adjusted real GDP growth, as a proxy for expectations
  - positive and significant

### Debt overhang / creditworthiness (supply-side)
- Montenegro
  - NPL ratio, six month average, lagged
  - negative and significant
- Lithuania
  - NPL ratio (aggregate), one month lagged
  - negative and significant
- Latvia
  - NPL ratio for corporate loans, monthly data since 2008, quarterly data were used before 2008 and interpolated
  - negative and significant
- Poland
  - NPL ratio for corporate loans, one month lagged
  - negative and significant
- Romania
  - share of overdue loans in total lending
  - negative and significant

### Funding cost / financial stress (supply-side)
- Montenegro
  - lending rate margin over Euribor, percent change
  - negative, insignificant
- Lithuania
  - interest margin, calculated as lending rate minus funding costs. Funding costs calculated as the weighted average of litas and euro deposit rate and cost of external funding; several measures were not significant and were dropped
- Latvia
  - deposit rate on zloty deposits, in real terms, deflated by HICP
  - negative and insignificant
- Poland
  - deposit rate weighted average by currency
  - negative and significant
- Romania
  - cost of external funding from parent banks is proxied by 12m EURIBOR rate plus CDS of Sweden and Lithuania
  - positive and significant

### Lending capacity (supply-side)
- Montenegro
  - domestic deposits, six month average, deflated, lagged
  - positive and significant
  - Banks’ foreign liabilities, six month average, deflated, lagged, instrumented with sovereign CDS spreads
  - positive and significant
- Lithuania
  - domestic deposits, lagged, seasonally adjusted series, in logs
  - positive and significant
  - parent funding, lagged one month (First order difference in parent funding was instrumented with the CDS spread of SEB bank; the fitted value of stock of parent funding (backed up using the fitted values for difference in parent funding) was included in the supply function), in real terms, deflated by HICP, in logs, seasonally adjusted series
  - positive and significant
- Latvia
  - Deposits deflated by HICP
  - Positive and insignificant
  - The change in parent funding was instrumented with the CDS spread of SEB bank; the moving average over the last 3 months of fitted change in parent funding deflated by HICP was included in the supply equation
  - negative and insignificant
  - The CDS spreads for SEB bank was used as a proxy for funding costs of parent banks as it goes back to 2004 and is highly correlated with CDS of other parent banks such as Swedbank or Nordea.
- Poland
  - domestic deposits minus banks’ reserves at the Polish central bank, in real terms, deflated by HICP, in logs
  - positive and significant
- Romania
  - bank capital divided by minimum capital requirements, in real terms
  - positive and significant
  - (Different measures of flows from BIS reporting banks to the CEE region as a proxy for availability of external funding was tried, but was not significant in most specifications and hence dropped)

### Other supply-side variables / dummies
- Montenegro
  - Dummy =1 from 2011 M11 onward to capture the bankruptcy of Snoras bank, and hence, its removal from the statistics
- Lithuania
  - lending survey on the percentage of respondents that identify constraints to obtaining financing as a significant factor limiting production
  - Negative and significant

### Household supply — return, economic conditions, debt overhang, lending capacity
- Return of credit (households)
  - Montenegro: Lending rate on new loans (in percent), constructed as the weighted average of housing and other loans in lats and in euros, real terms, deflated by HICP — positive and significant
  - Lithuania: lending rate on new zloty household credit, in real terms, deflated by HICP — positive and significant
  - Latvia: lending rate, average, nominal — positive and significant (inflation was insignificant and hence dropped)
- Economic conditions (households)
  - Montenegro: stock exchange index — positive and significant
  - Lithuania: expected business situation in the retail sector, seasonally adjusted — positive and significant
  - Poland: linearly interpolated seasonally adjusted real GDP growth, as a proxy for expectations — positive and significant; other variables produced a poorer fit and were dropped
- Debt overhang / creditworthiness (households)
  - Montenegro: NPL ratio for households, monthly data since September 2008, before that quarterly data were interpolated — negative and significant
  - Lithuania: NPL ratio on household loans, one month lagged — negative and significant
  - Latvia: share of overdue loans in total — negative and significant
- Lending capacity (households)
  - Montenegro: deposits deflated by HICP — positive and significant
  - Lithuania: The change in parent funding was instrumented with the CDS spread of SEB bank; the moving average over the last 3 months of fitted change in parent funding deflated by HICP was included in the supply equation — positive and insignificant
  - Poland: domestic deposits minus banks’ reserves at the central bank, 4-month lagged, in real terms, deflated by HICP, in logs — positive and significant
  - Romania: external funding, lagged one month, first order difference in external funding was instrumented with the CDS spread of parent banks, then the fitted value of stock of parent funding (backed up using the fitted values for difference in external funding) was included in the supply function — positive and significant
  - Romania: bank capital divided by minimum capital requirements, in real terms — positive and significant
  - Romania: flows from BIS reporting banks to countries in the CEE region as a proxy for availability of external funding (percent change in flows in US$) — positive and significant

### Funding costs / financial stress (households)
- Montenegro
  - real deposit rate on new deposits — negative and significant
  - other indicators, such as WIBOR-OIS spread and CIRS rate turn out to be insignificant
- Lithuania
  - deposit rate, average, nominal — negative and significant

### Sample period by country
- Montenegro: 2007M1–2012M12
- Lithuania: 2006M10–2012M10
- Latvia: 2004M12–2012M9
- Poland: 2005M12–2012M9
- Romania: 2005M1–2012M8

### Appendix IV. Fit of Model (figures overview)
- Note: While the models are usually estimated using data on real credit, the figures in Appendix IV report actual and fitted credit in nominal terms.
- Montenegro: Actual and Predicted Credit (In millions of Euro) — plotted Jan-07 through Oct-12; series shown: Predicted credit, Actual credit
- Lithuania: Actual and Predicted Credit (In millions of litai) — plotted Oct-06 through Aug-12; series shown: Predicted credit, Actual credit; Sources: Central Bank of Lithuania; Haver; and IMF staff estimates.
- Latvia: Households and NFC Actual and Predicted Credit (In millions of lats) — plotted Dec-04 through Jun-12; series shown: Predicted credit, Actual credit; Sources: Haver; and IMF staff estimates.
- Poland: Households and NFC Actual and Predicted Credit (In millions of zloty) — plotted Dec-05 through Jun-12; series shown: Predicted credit, Actual credit; Sources: Haver; and IMF staff estimates.
- Romania: Households and NFC Actual and Predicted Credit (In millions of RON) — plotted Mar-05 through Aug-12 for households and Jun-05 through Jul-12 for NFC; series shown: Predicted credit, Actual credit; Sources: Haver; and IMF staff estimates.

### Appendix V. Robustness Check on Significance of Excess Supply (figures overview)
- Montenegro & Lithuania: Excess Supply (In millions of euro / litai) — plotted series Jan-07 through Sep-12 (Montenegro) and Oct-06 through Aug-12 (Lithuania) with (+ 2std, -2std) bands; series shown: Excess supply
- Latvia: Households and NFC Excess Supply (In millions of lats) — plotted Dec-04 through Jun-12 for households and Dec-04 through Jun-12 for NFC with (+2 std, -2 std) bands; series shown: Excess Supply
- Poland: Households Excess Supply (In million of zloty) and NFC Excess Supply (In millions of zloty) — plotted Dec-05 through Aug-12 with (+ 2 stdev, -2 stdev) bands; series shown: Excess supply
- Romania: Households and NFC Excess Supply (In millions of RON) — plotted Jun-05 through Jun-12 with (+2 std, -2 std) bands; series shown: Excess Supply
- Sources for robustness figures: Haver; and IMF staff estimates.

*Source: _wp1515 - Appendix III. Description of Variables of Disequilibrium Model (IMF staff content).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1515.pdf_